Optimization of feature selection in face recognition system using differential evolution and genetic algorithm

Abstract

This paper compares the results of the optimization techniques for feature selection of face recognition system in which face as a biometric template gives a large domain of features for optimizing feature selection. We attempt to minimize the number of features necessary for recognition while increasing the recognition accuracy. It presents the application of differential evolution and genetic algorithm for feature subset selection. We are using local directional pattern (LDP), an extended approach of local binary patterns (LBP), to extract features. Then, the results of DE and GA are compared with the help of an extension of support vector machine (SVM) which works for multiple classes. It is used for classification. The work is performed on 10 images of ORL database resulting in better performance of differential evolution.